Production experience
Architected distributed infrastructure processing billions of records and millions of queries a day at better than 99.9% uptime.
Production systems, not just prototypes.
Read the resumeServices
More than fifteen years architecting production systems at scale, from government and enterprise data infrastructure to garage-lab instruments. Same rigor, both rooms.
Five practices, one practitioner
Each engagement draws on one or more of these. Nothing is bolted on afterwards. Security, monitoring, and a clean way to retire the system come standard.
Distributed systems
Systems that move billions of records with high availability and fault tolerance.
Distributed databasesMessage queuesLoad balancingFault tolerance
Worked examples
- Search infrastructure at TransUnion scale
- High-volume messaging platforms
- Multi-region data replication
Data engineering
Pipelines that ingest, transform and serve, real time and batch, without a cloud bill.
ETLWarehousingStream processingAnalytics
Worked examples
- 4.9B data point integration
- Snowflake data architectures
- Real-time reporting backends
API design
REST and WebSocket APIs with auth, rate limiting and monitoring from the first commit.
FastAPIRESTWebSocketGraphQL
Worked examples
- 85-endpoint API orchestration
- Multi-carrier messaging integrations
- Real-time event streaming
Proof you can open
- TerraPulse’s MCP serverPublic and read-only: an AI agent can query the measured data directly, with no key. (another site, opens in a new tab)
- The 6502 APIThe transistor-level 6502 as a web service, one half-cycle at a time. Each request carries the whole machine, and AI assistants can call it too. (another site, opens in a new tab)
Edge and IoT
AI and data on small, low-power devices: custom protocols, wireless sensor networks, salvaged hardware.
Raspberry PiCustom protocolsMesh networksLow power
Worked examples
- 60-node Pi cluster
- Custom 802.11 protocols
- LoRa mesh deployments
AI integration
Machine learning in production: answers in under a tenth of a second, several AI providers working together, and the safeguards around them.
LLM integrationModel deploymentVector searchML pipelines
Worked examples
- Multi-provider AI orchestration
- Sub-100ms inference pipelines
- Self-improving systems
Proof you can open
Filled: reporting right now.Hollow: quiet or offline, but the page still opens. What each one reads
Three shapes, and what they cost
Most engagements start as a short scoping call, then drop into one of the shapes below. Hybrid arrangements are fine. Say so up front.
Project
$25K to $100K
Fixed scope, defined deliverables
Best forSpecific initiatives with clear requirements
Hourly
$150 to $275/hr
Flexible, as-needed expertise
Best forTechnical guidance and architecture review
Retainer
$15K to $50K/mo
Dedicated monthly support
Best forOngoing partnership and continuous improvement
Four steps, no theatre
The same rhythm every time. Discovery is unbillable. If the work doesn’t fit, I’ll say so before we sign anything.
| Step | What happens |
|---|---|
| 1Discovery | Understand the problem deeply before proposing solutions.Unbillable |
| 2Design | Architecture that balances elegance with practicality. |
| 3Build | Iterative development with continuous feedback. |
| 4Deliver | Documentation, handoff, and knowledge transfer. |
What you get that doesn’t come standard
Government and classified work
Lead developer on a classified project with the United States Government (TransUnion TruLookup, 2014 to 2018).
Extensive experience in secure environments handling sensitive investigative and personal data for law enforcement and federal agencies.
If your work carries a clearance requirement, say so in the first email.
The section in the resumeMaker mentality
Creative problem-solving, from building a 60-computer Raspberry Pi cluster out of salvaged hardware. In the lab now: a 6502 chip simulated transistor by transistor from photographs of its silicon, and an NES console running fast models of its two chips, each checked against a full simulation. Lab work, not for sale.
tinymachines.ai (another site, opens in a new tab)Full stack
From low-level protocols to cloud architecture to ML pipelines. One head, one accountability.
The work recordFour of the proofs above report live
What each one is doing as you read, asked again every 15 seconds.
| Claude Code | 14 active min since 14:00 UTC | last heard |
| Knock knock | 8,541 probes trapped since 00:00 UTC | last heard |
| Hotbits | no reading | checking |
| Perception bus | 16.5 events/s | last heard |
As of
/api/now: each row read from its own instrument, held 10 s on the server
Questions I get
Anti-cloud means host-local, and that goes for the practice too. I’m based in Grand Rapids, Michigan and work with teams across Ada, Cascade, East Grand Rapids, Kentwood, and the wider Kent County. On site when it helps, remote when it doesn’t.
Do you provide AI and data-engineering consulting in Grand Rapids?
Yes. I'm based in Grand Rapids, Michigan and work with teams across Kent County and West Michigan, on-site or remote, on AI integration, data pipelines, and production systems at scale.
What is edge computing, and can you build it for a West Michigan business?
Edge computing runs data processing on local hardware instead of a distant cloud, cutting latency and cloud bills. I design and build edge and IoT systems for businesses in Grand Rapids, Ada, Cascade, and the wider Kent County area.
Do you work on-site in the Grand Rapids area?
Yes. As a Grand Rapids-based technologist I can work on-site across Kent County when it helps, and remotely for everything else.
Let’s sketch it
Have a project in mind? Let’s talk about how I can help. A first email can be one paragraph. The contact page lists the four things that make it easy to answer.